Offline reinforcement learning (RL) enables effective learning from previously collected data without exploration, which shows great promise in real-world applications when exploration is expensive or even infeasible. The discount factor, $γ$, plays a vital role in improving online RL sample efficiency and estimation accuracy, but the role of the discount factor in offline RL is not well explored. This paper examines two distinct effects of $γ$ in offline RL with theoretical analysis, namely the regularization effect and the pessimism effect. On the one hand, $γ$ is a regulator to trade-off optimality with sample efficiency upon existing offline techniques. On the other hand, lower guidance $γ$ can also be seen as a way of pessimism where we optimize the policy's performance in the worst possible models. We empirically verify the above theoretical observation with tabular MDPs and standard D4RL tasks. The results show that the discount factor plays an essential role in the performance of offline RL algorithms, both under small data regimes upon existing offline methods and in large data regimes without other conservatisms.

With the increasing presence of robots in our every-day environments, improving their social skills is of utmost importance. Nonetheless, social robotics still faces many challenges. One bottleneck is that robotic behaviors need to be often adapted as social norms depend strongly on the environment. For example, a robot should navigate more carefully around patients in a hospital compared to workers in an office. In this work, we investigate meta-reinforcement learning (meta-RL) as a potential solution. Here, robot behaviors are learned via reinforcement learning where a reward function needs to be chosen so that the robot learns an appropriate behavior for a given environment. We propose to use a variational meta-RL procedure that quickly adapts the robots' behavior to new reward functions. As a result, given a new environment different reward functions can be quickly evaluated and an appropriate one selected. The procedure learns a vectorized representation for reward functions and a meta-policy that can be conditioned on such a representation. Given observations from a new reward function, the procedure identifies its representation and conditions the meta-policy to it. While investigating the procedures' capabilities, we realized that it suffers from posterior collapse where only a subset of the dimensions in the representation encode useful information resulting in a reduced performance. Our second contribution, a radial basis function (RBF) layer, partially mitigates this negative effect. The RBF layer lifts the representation to a higher dimensional space, which is more easily exploitable for the meta-policy. We demonstrate the interest of the RBF layer and the usage of meta-RL for social robotics on four robotic simulation tasks.

Experience replay methods, which are an essential part of reinforcement learning(RL) algorithms, are designed to mitigate spurious correlations and biases while learning from temporally dependent data. Roughly speaking, these methods allow us to draw batched data from a large buffer such that these temporal correlations do not hinder the performance of descent algorithms. In this experimental work, we consider the recently developed and theoretically rigorous reverse experience replay (RER), which has been shown to remove such spurious biases in simplified theoretical settings. We combine RER with optimistic experience replay (OER) to obtain RER++, which is stable under neural function approximation. We show via experiments that this has a better performance than techniques like prioritized experience replay (PER) on various tasks, with a significantly smaller computational complexity. It is well known in the RL literature that choosing examples greedily with the largest TD error (as in OER) or forming mini-batches with consecutive data points (as in RER) leads to poor performance. However, our method, which combines these techniques, works very well.

We present an empirical study on the use of continual learning (CL) methods in a reinforcement learning (RL) scenario, which, to the best of our knowledge, has not been described before. CL is a very active recent research topic concerned with machine learning under non-stationary data distributions. Although this naturally applies to RL, the use of dedicated CL methods is still uncommon. This may be due to the fact that CL methods often assume a decomposition of CL problems into disjoint sub-tasks of stationary distribution, that the onset of these sub-tasks is known, and that sub-tasks are non-contradictory. In this study, we perform an empirical comparison of selected CL methods in a RL problem where a physically simulated robot must follow a racetrack by vision. In order to make CL methods applicable, we restrict the RL setting and introduce non-conflicting subtasks of known onset, which are however not disjoint and whose distribution, from the learner's point of view, is still non-stationary. Our results show that dedicated CL methods can significantly improve learning when compared to the baseline technique of "experience replay".

This paper makes progress towards learning Nash equilibria in two-player zero-sum Markov games from offline data. Specifically, consider a $\gamma$-discounted infinite-horizon Markov game with $S$ states, where the max-player has $A$ actions and the min-player has $B$ actions. We propose a pessimistic model-based algorithm with Bernstein-style lower confidence bounds -- called VI-LCB-Game -- that provably finds an $\varepsilon$-approximate Nash equilibrium with a sample complexity no larger than $\frac{C_{\mathsf{clipped}}^{\star}S(A+B)}{(1-\gamma)^{3}\varepsilon^{2}}$ (up to some log factor). Here, $C_{\mathsf{clipped}}^{\star}$ is some unilateral clipped concentrability coefficient that reflects the coverage and distribution shift of the available data (vis-\`a-vis the target data), and the target accuracy $\varepsilon$ can be any value within $\big(0,\frac{1}{1-\gamma}\big]$. Our sample complexity bound strengthens prior art by a factor of $\min\{A,B\}$, achieving minimax optimality for the entire $\varepsilon$-range. An appealing feature of our result lies in algorithmic simplicity, which reveals the unnecessity of variance reduction and sample splitting in achieving sample optimality.

The cybersecurity threat landscape has lately become overly complex. Threat actors leverage weaknesses in the network and endpoint security in a very coordinated manner to perpetuate sophisticated attacks that could bring down the entire network and many critical hosts in the network. Increasingly advanced deep and machine learning-based solutions have been used in threat detection and protection. The application of these techniques has been reviewed well in the scientific literature. Deep Reinforcement Learning has shown great promise in developing AI-based solutions for areas that had earlier required advanced human cognizance. Different techniques and algorithms under deep reinforcement learning have shown great promise in applications ranging from games to industrial processes, where it is claimed to augment systems with general AI capabilities. These algorithms have recently also been used in cybersecurity, especially in threat detection and endpoint protection, where these are showing state-of-the-art results. Unlike supervised machines and deep learning, deep reinforcement learning is used in more diverse ways and is empowering many innovative applications in the threat defense landscape. However, there does not exist any comprehensive review of these unique applications and accomplishments. Therefore, in this paper, we intend to fill this gap and provide a comprehensive review of the different applications of deep reinforcement learning in cybersecurity threat detection and protection.

As companies look for ways to put their data to use, reinforcement learning (RL) is becoming an increasingly inviting and accessible option. General purpose RL tools – such as Microsoft's Project Bonsai – are now available, waiting to be utilized for planning, optimization, and automation. Between the promises of next-gen AI and the availability of tools in the cloud, it seems like the market is building up towards an explosion of RL-powered business solutions. The remainder of this article assumes the reader is familiar with the concepts of RL. If not, we recommend starting with this article. Reinforcement learning has had some well publicized successes over the past decade that have piqued the interest of businesses.

To tackle the heterogeneous requirements of beyond 5G (B5G) and future 6G wireless networks, conventional medium access control (MAC) procedures need to evolve to enable base stations (BSs) and user equipments (UEs) to automatically learn innovative MAC protocols catering to extremely diverse services. This topic has received significant attention, and several reinforcement learning (RL) algorithms, in which BSs and UEs are cast as agents, are available with the aim of learning a communication policy based on agents' local observations. However, current approaches are typically overfitted to the environment they are trained in, and lack robustness against unseen conditions, failing to generalize in different environments. To overcome this problem, in this work, instead of learning a policy in the high dimensional and redundant observation space, we leverage the concept of observation abstraction (OA) rooted in extracting useful information from the environment. This in turn allows learning communication protocols that are more robust and with much better generalization capabilities than current baselines. To learn the abstracted information from observations, we propose an architecture based on autoencoder (AE) and imbue it into a multi-agent proximal policy optimization (MAPPO) framework. Simulation results corroborate the effectiveness of leveraging abstraction when learning protocols by generalizing across environments, in terms of number of UEs, number of data packets to transmit, and channel conditions.

Let's start our journey by looking directly in the lion's den (Algorithm 1 below). It looks pretty scary at first sigh… But no worries, we will go through it line by line and unpack each concept. First things first, in Deep Q-Network we have a Q. And, in the beginning Sutton & Barto created the Q-function. The Q- function (a.k.a. the state-action value function), maps a couple of state, action to a value (or expected reward).

With the development of artificial intelligence,more and more financial practitioners apply deep reinforcement learning to financial trading strategies.However,It is difficult to extract accurate features due to the characteristics of considerable noise,highly non-stationary,and non-linearity of single-scale time series,which makes it hard to obtain high returns.In this paper,we extract a multi-scale feature matrix on multiple time scales of financial time series,according to the classic financial theory-Chan Theory,and put forward to an approach of multi-scale stroke deep deterministic policy gradient reinforcement learning model(MSSDDPG)to search for the optimal trading strategy.We carried out experiments on the datasets of the Dow Jones,S&P 500 of U.S. stocks, and China's CSI 300,SSE Composite,evaluate the performance of our approach compared with turtle trading strategy, Deep Q-learning(DQN)reinforcement learning strategy,and deep deterministic policy gradient (DDPG) reinforcement learning strategy.The result shows that our approach gets the best performance in China CSI 300,SSE Composite,and get an outstanding result in Dow Jones,S&P 500 of U.S.